NEMD M-P thermal conductivity equation

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps users,

Hi, I am trying to calculate thermal conductivity of silicon nanowires using Muller-Pathe NEMD simulation. I read through the paper by Muller-Pathe and know that thermal conductivity is calculated by the ratio of energy flux to the temperature gradient, which is described in the paper as (energy exchange/(2 x timestep x Area of heat transfer x T gradient)). However, I am having hard time actually utilizing this equation in my input file, and getting very high thermal conductivity (several hundreds of thousands). I searched around the forum and found that some people actually implemented the formula as {energy exchange/((2 or 4 sometimes) x timestep x (10 or 100) x Area of heat transfer x T gradient)}. I wonder why in the denominator there sometimes is 4 instead of 2, and 10 or 100. Here are the links of those threads.

http://lammps.sandia.gov/threads/msg36465.html

http://lammps.sandia.gov/threads/msg11716.html

Thanks in advance,

2

The examples/KAPPA dir has scripts that use the MP (and other methods)
for thermal cond. You can take a look at those.

I will say (from painful experience) that you have to worry about factors
of 2 (e.g. heat flowing both directions for a periodic system) and cross sectional
areas if you are in LJ units can be non-obvious.

Steve